Aspects, rep specs, shifts, conditional expressions, expression functions

Aspects, rep specs, shifts, conditional expressions, expression functions

[Bill Findlay]

The topics listed in the subject have been controversial recently.
I wonder what people think of my first attempt at an Ada 2012 package,
which seems to use them all, in what I hope is a judicious mix?

----
with Unchecked_Conversion;
with KDF9.CPU;

package KDF9.K5_stuff is

   -- PHU, the Program Hold-Up register is internal to I/O Control.
   -- A subset is exposed to Director by means of the K5 order.
   -- It has one element for each of the 4 program priority levels.
      
   type blockage is (buffer_busy, locked_out) with Size => 1;
   
   type group_address is mod 1024 with Size => 10;
    
   type PHU_store (hold_up : one_bit := 0;
                   reason  : blockage := buffer_busy) is
      record
         case hold_up is
            when 0 =>
               null;
            when 1 =>
               case reason is
                  when buffer_busy =>
                     buffer_nr  : buffer_number;
                  when locked_out =>
                     group_nr   : group_address;
               end case;
         end case;
      end record;

   PHU : array (priority) of PHU_store;

   type PHU_subset is
      record
         hold_up   : one_bit;
         reason    : blockage;
         parameter : buffer_number;
      end record
   with Size => 6, Bit_Order => Low_Order_First;
   
   for PHU_subset use
      record
         hold_up   at 0 range 5 .. 5;
         reason    at 0 range 4 .. 4;
         parameter at 0 range 0 .. 3;
      end record;
      
   type PHU_as_6_bits is mod 2**6 with Size => 6;

   function as_6_bits is
      new Unchecked_Conversion(Source => PHU_subset,
                               Target => PHU_as_6_bits);

   function K5_word (short_PHU : PHU_subset) return word is
      (word(as_6_bits(short_PHU)));

   function short_PHU (p : priority) return word is
      (K5_word((
                hold_up   => PHU(p).hold_up,
                reason    => (
                              if PHU(p).hold_up = 0 then
                                 buffer_busy  -- A non-significant 0 bit.
                              else
                                 PHU(p).reason
                             ),
                parameter => (
                              if PHU(p).hold_up = 0 then
                                 0
                              elsif PHU(p).reason = buffer_busy then
                                 PHU(p).buffer_nr
                              else
                                 buffer_number(PHU(p).group_nr mod 2**4)
                             )
              ))
      );

   -- A K5_operand is a KDF9 word, D00-D47, with the content:
   --    short_PHU(0) in D00 .. D05
   --    short_PHU(1) in D06 .. D11
   --    short_PHU(2) in D12 .. D17
   --    short_PHU(3) in D18 .. D23

   function K5_operand return word is
      (
       KDF9.CPU.shift_word_left(short_PHU(0), 47-05) or
       KDF9.CPU.shift_word_left(short_PHU(1), 47-11) or
       KDF9.CPU.shift_word_left(short_PHU(2), 47-17) or
       KDF9.CPU.shift_word_left(short_PHU(3), 47-23)
      );

end KDF9.K5_stuff;

-- 
Bill Findlay
with blueyonder.co.uk;
use  surname & forename;

Re: Aspects, rep specs, shifts, conditional expressions, expression functions

[Randy Brukardt]

"Bill Findlay" <yaldnif.w@blueyonder.co.uk> wrote in message 
news:CB186063.D1C0%yaldnif.w@blueyonder.co.uk...
...
>   type blockage is (buffer_busy, locked_out) with Size => 1;

Minor stylistic comment:
The intended style for aspect_specifications is to have them indented on the 
line following the declaration, as follows:

   type blockage is (buffer_busy, locked_out)
      with Size => 1;

We had a huge argument about this in the ARG. Note that the recommended 
indentation style for any Ada construct is that shown in the syntax diagrams 
in the RM. (The layout of those productions is not random, even if a lot of 
people think it is!)

Of course, you can do whatever you want -- just keep in mind that there is 
an intended style for Ada.

                                  Randy. 

Re: Aspects, rep specs, shifts, conditional expressions, expression functions

[Bill Findlay]

On 23/12/2011 01:01, in article jd0jtn$21o$1@munin.nbi.dk, "Randy Brukardt"
<randy@rrsoftware.com> wrote:

> "Bill Findlay" <yaldnif.w@blueyonder.co.uk> wrote in message
> news:CB186063.D1C0%yaldnif.w@blueyonder.co.uk...
> ...
> 
>>   type blockage is (buffer_busy, locked_out) with Size => 1;
> 
> Minor stylistic comment:

Thanks, Randy.  I had not realized that the RM authors were setting a
precedent with the layout.

One aspect (!)  of my code thatI do not like is that the auxiliary type,
PHU_as_6_bits, and functions, as_6_bits and short_PHU, are exposed to view.
I'm going back to a more conventional package with a body.

-- 
Bill Findlay
with blueyonder.co.uk;
use  surname & forename;